Is anyone actually utilising a recent version of PostgreSQL for full-text searching beyond a hobby project? How do you find the speed and accuracy versus Elasticsearch?
I've also used Elasticsearch, and I reckon that's pretty damn amazing.
Anyone wanting more in-depth information should read or watch this FTS presentation from last year. It's by some of the people who has done a lot of work on the implementation, and talks about 9.6 improvements, current problems, and things we might expect to see in version 10. https://www.pgcon.org/2016/schedule/events/926.en.html
There's also some previous presentations on the same topic which are interesting. You can see the RUM index (which has faster ranking here): https://github.com/postgrespro/rum
For, say, a tag-based search system (think any forum or booru-style imageboard), is Elasticsearch completely overkill?
I realise I could probably just google it™, but it's a lot easier to understand a product's strengths when they're put in a situational context.
Solr had more functionality (largely by being able to throw more nodes at the problem), but for our purposes, PostgreSQL's fts was certainly good enough. There were a few cases where we had to be careful so the full text index would be used (because data distribution did not always match the planner's assumptions) but on the whole it worked very well.
I was using Django and rewrote the search just simply collecting a dictionary of terms that I would then explode into filter on the model.
A database of order 100k articles with something like 50 fields, about 20 of them foreign keys, bulk text fields of several to tens of KB responded within milliseconds on a 4 core 4GB machine under load.
The "weird thing" about it that I didn't understand at the time was that using a more complex query would often result in speedups. I understand it better now after this article since PG builds some sort of index on text fields.
PG is one of the infrastructure pieces that you install and then it just runs performantly without much fuss. There probably are cases where it fails horribly but in almost a decade of use I have yet to see a single operations failure.